HUFFMAN CODING
Ms 140400147
SADIA YUNAS BUTT
Abstract
In 1952 David A.Huffman the student of
MIT discover this algorithm during work on
his term paper assigned by his professor
Robert M.fano.The idea came in to his mind
that using a frequency sorted binary tree we
can able to find out the most efficient binary
code[1].Huffman deny the major flaw of
Claude Shannon Fano coding by building
the tree from bottom up instead of from the
top down .There is a lot of compression
algorithm but we focus on lossless
compression so in this regard Huffman
algorithm is so cool and efficient algorithm
in sort of smallest application .Huffman use
for image compression for example PNG
,JPG for simple picture of bird it also use
MPEG whenever you watch movie or video
in computer chances are that its image will
be in MPEG format both these PNG MPEG
and for text compression for example static
and dynamic Huffman coding techniques
INTRODUCTION
Huffman coding is a great way to compress
down and to able to shrink it down by bites
and chunks so give you example
how it works write down on right an side
the letter A B C D E if the document only
have a range of these five letters and we
want to do compress the data Popping could
be the great way for doing it and the reason
why ? if we want to stress the data in regular
text format usually with store data you will
after do like this you need set of your system
and each one would have to represent a bit is
Use of Huffman coding method to compress
their content most of the time images and
movies are also compress using lossless
technique.so it is general technique which
we often use in our computer like PKZIP in
which WINZIP is one and other is GZIP
these are use for general purpose algorithm
which compress your text file and these use
Huffman another application of Huffman is
audio so there is MP3 and AAC format for
audio whenever we use Ipod we use MP3 all
these are encoded compress using Huffman
there is lossly compression but there is
lossless compression as well AAC is very
widely use in U tube so rightly now we are
using Huffman algorithm in your Iphone in
Ipads these are use AAC .Huffman proof
that compression can be easy using a
frequency sorted binary tree in the bottom
up fashion.
a two row at each -and so u say it you have
to take a 4 bit for every letter but there is
actually a lie
u looking have been 8 bit for every letter
but that is actually wrong, a better way for
doing it and that was Huffman coding does
and its only work on specific data so if you
need 4 range data this is the something that
would work for you but we can do actually after sum up it root node contain which does
is take care similar to something like this not contain any node a sum of total data we
we can actually take care and stated looking for
acquiring two row 8 after you can do that
just by using 3 bits and what can do have to
do it in 3 form first then i am assure you
have actual code soft works very first we

why it will be better then ASCII

can do is . if you wondering when which
actually is important certain strategy in
computer science paring matrix that this is
actually critical because you are doing with
dynamic command of data and you need to
be compress and Huffman coding is best
way to doing that so the temporary tree
generate is can be robed different then some
the other we have seen as specifically can
be different like binary tree because it
cannot be use c associated with them and
we can actually stood of with two lowest
values we can give a E and we say it is
equal to 8 e:8 and we draw a box
around it
and we know that D is the same thing and
we say they are equal 16
we will assign here the values 0 at the left
hand side and 1 at the right hand side
we see the value 0 to A and if we see at
the binary tree we will see loop here from
root to B we will see binary value 100 and in
the same way the all nodes get there values
now we store all this in compress form
instead of 8 value here we give 1 and next
we see three give value 8 same way do this
all we are doing is the shrinking down the
numbers of bits that required
if we look here is 5 time 8 so we need a total
40 as compare the new system Huffman
coding
which is total 1+ 4 set of three so we
looking 13

starting with the low value first so we can

start of c:10 B:12 so they will be equal to
22 and now we connect it 22 and 16 it will
make 38 and then we connect it to A :24
 by Huffman coding we see well over half of
the amount of space and that's one of the big
reason that Huffman coding so powerful
because by simply doing something by
shrinking doing that are necessary on per
letter base they much more for sequences or
something where u can were the data sort
like that
In this paper we may discuss first that what
is Huffman coding and then how this
algorithm works In 3 section we will see
adoptive Huffman coding how Huffman
encode decode , image text compression
finally we will discuss the conclusion and
the future scope of it.
WHAT IS HUFFMAN CODING
There are many ways to store information.
computer scientist always looking for new
and better ways to store strings of data with
as little space as possible. Huffman coding is
a method of storing strings of data as binary
code in an efficient manner .Huffman
coding uses" lossless data compression"
which means no information is lost which
you are coded[3]
Huffman coding uses ' variable length [6]
coding' which means that symbol in the data
you are encoded or converted to binary
symbol based on how often that symbol is
used .for example if character 'a ' is used in
your data a lot , the binary symbol
representing it is shorter. If it is used rarely
the symbol representing it is longer .this way
all the data will be take less physical space
when encoded. There is a way to decide
what binary code to give each character
using trees.
char frequency
A 1
B 6
C 7
D 2
E 8
in this figure we have 5 different letter and
also have frequency which often be use a
and d in this example is the least ,
frequency represent how often character
appear in a string of data now imagine these
as 5 separate trees combine the two smallest
tress in order we can combine them slowly
bit by bit first we combine a and d
which are smallest after combining them i
have new tree with a greater frequency 3
now i will go for next least number which is
here b:6 now i have new tree with greater
frequency 9. next i have two more least
character c and e and get the frequency
15 .finally i combine these two tress and get
one large tree .now we have large tree
containing all the characters we can now
assign them a binary code to each symbol by
going down the tree (each left branch
receives a '0' and each right branch will
receive '1' .if we go to the top we will get a
binary string a=000,b=01 , c=10 , d=001
These are what a ,b,c,d,e will each be
converted to.
Huffman coding is to encode data to take up
less space wouldn't it make sense to give
some character binary code only 1 digit long
('0' or'1')or shouldn't more character be
given code 2 digit long instead of 3?consider
how you would read the code it is important
for how each representation to be unique
from each other such that you can easily tell
which character that part is suppose to each
represent .If '1' and '0' is representing a
character any other representation
containing 1 and 0 could be different
character . now encode a,b,c,d,e using
Huffman code so we will take each
character in replace it with the binary
abcbe=000 01 1001 11-00001100111
,e=11
if we have to decode it you have to start
either from back way front or front
suppose we have to accrue the average
length via Huffman coding there is special
formula for that so average code word
length=1/f(T)*sum of d(i)*f(i), for i=0 to
i=n
where n = number of character
f(T)=total frequency
f(i)=frequency of that character
d(i)= length of that symbol
if we consider it in our example[2]
Average length
=(1/(1+6+7+2=8)*(3*1+2*6+2*7+3*2+2*8
)

decode 1011001000011101 =(1/24)*(3+12+14+6+16)

we have to compare the representation =(1/24)*(51)

above to the binary code bit by bit to fill the =2.125 digit long
only possible result
1011001000011101=c1001000011101(only (the average letter in the cod)[8]
c start with 1,then 0) so it must start with 'c'
after that it will be start with 'e'=c e
001000011101(only e start with 1 ,then 1)
then start with =c e d 000011101(only d
start with 0 then 0 then 1)
= c d e a 011101(only a
start with 0 then 0 then 0)
=c d e a b1101(only b start
with 0 then 1)
=c d e a b e 01(only e start
with 1 then 1)
=c d e a b e b=cdeabeb
e is encoded to 2.125 binary digit.

Adoptive Huffman coding

In adoptive Huffman coding we should
remember two steps are . nodes should be
increasing order and of the frequency and
sum of the frequency in parent node greater
than all the frequency of the side node .so
taking these two facts in count it grow the
Huffman tree using the adoptive Huffman
coding [3],[4]
now we will see all left hand side nodes are
First of all in the for the string greater than right hand side so in this case
(MISSISSIPPI) we write an empty node we will see that it is not in proper order so
always we first insert empty node after
doing this we insert m we will re-loop this node( )

so we put two empty node here and put the

frequency of m here that is m:1 we should
always update the parent node so 1+0 =1
now we insert i so again we put two empty
we go back and interchange this node .[5]
node and insert i in it and the frequency of
i:1 and update its parent node which was so in the string (MISSISSIPPI) MISS
empty so after putting it becomes 1 comes on right hand side and PPI go to left

now we can put s so we again make two

empty node and insert s in the right node
and insert the frequency of s which is s:1
and update its parent node we also update
the i parent node which is now 1+1=2
side) after interchange we will get this tree
after each update in parent node we omit the
previous frequency and write new after add
in the previous.
now we have to add next s but there is

frequency in empty node so update here the

parent node
already s so we just update the frequency
but to do this we will see problem with
updating see what will happened when we 1
more s in the previous s frequency
so we will interchange this 1:s with 1:m

but here we will see 5 > 4 so we will

interchange them with each other because
they should be either less or equal to the
right hand side.
so then we continue the operation s is
already here so we update it and it becomes
2 now it look healthy tree so now we add
new frequency of I here but there is
already I so we just update the I so it will
become 2 and if go to remaining parent node
they all will add 1 more in them see
now we take next s from the string and
update it ,now it becomes 3 now we do
interchange these two nodes so s frequency
will go to right side.So this is the update tree
next is again frequency i so we add new
now we again insert next p and if we see
string it in it will be now 3 then we take
here we have p so we update the
string p there is no p that why we put this
frequency it become 2 and its parent node
become 1+1=3 and their parent node will
 so due to this interchange whole tree will
change the nodes frequency now it become
healthy tree .now we add last character i
we have i so we update its frequency it will
become 4 then we update its parent node it
will become
if we check the integrating of the tree it is in

become 3+3=6 and lastly 6+4=10 at the root

node

increasing order so it is healthy tree all the

parent node are greater then its child node

Huffman encode decode [2]

sally sell sea shells
we put the binary values to this string
s 10
now if we look at the tree we will see it is
again not a healthy tree because 2> 1 so we a 1111
have to interchange it but now at this time
we will interchange 1:m with 2:p l 01
y 1110
_ 001
e 110
h 0001
! 0000
now we assign values to these characters
 then we add them and make the root node so
on. Now we write the branches as 0 and 1

Text compression With Huffman Coding

this is the complete construction of tree. To
Huffman coding technique use for
find the binary tree the code of the letter just
compression .there is a idea that more
start the top of the tree and follow tree
common letter equal fewer bits, and
branches encoding of I is 00 and for S is
similarly less common letter require more
bits to represent them[7].in this case we use 01 for P encoding is 100 do the same things
the string MISSISSIPPI for each other letter. We will see that letter
RIVER=17characters .If we reduce the with greatest frequency have the lower
presentation and lower frequency have the
common ASCII we required 8 bit each for
longest presentation that exactly we
total of 136 bit, Now we will write that how
wanted.[9]
much time each letter is present in the string.
We will SEE

Now we sort this string by order I5 S4 P2 now we write down the bits for message
R2 M1 V1 E1 -1.now we make a tree and
start with the small nods

this is Huffman coded binary message if we

count this message is equal to 46 bits this is
the lesser than the original 136 bits if we
divide them we get 33% of the original one
which save the 67 % of the original message
in this way we can compress the text
message through Huffman coding very
easily.
CONCLUSION
In this paper we proof that how Huffman is
cool and efficient algorithm in the filed of
compression although there are more
applications are available in the filed of
compression like arithmetic coding and
Lempel ziv, but mostly we see in our daily
life Huffman algorithm work in our Iphone
Ipods etc because it is not complex and more
optimal among the family of symbol code.
Huffman coding is a greedy algorithm so for
the compression of data it is consider good
that use longer bit instead of more frequent
bits but
Arithmetic coding gives greater
compression, is faster for adaptive models,
and clearly separates the model from the
channel encoding.
REFRENCES
[1].Gary Stix. Profile: David A. Huffman.
Scientific American, 265(3):54, 58,
September 1991.
[2]. Fundamentals of Information Theory
and Coding Design
By Roberto Togneri, Christopher J.S
deSilva
[3] Web Page Compression using
Huffman Coding Technique
Manjeet Gupta (Assistant professor)
Department of CSE JMIT Radaur
,Brijesh Kumar (associate professor)
Department of IT Lingyaya’s university.
[4] Fundamental Data Compression
Author(s):Ida Mengyi Pu
[5] A fast adoptive Huffman coding
algorithm
Communications, IEEE Transactions on
(Volume:41 , Issue: 4 )
[6] An efficient decoding technique for
Huffman codes. Rezaul Alam
Chowdhury(Department of Computer
Science and Engineering, Bangladesh
University of Engineering and Technology,
Dhaka-1000, Bangladesh)
Irwin King(Department of Computer
Science and Engineering, The Chinese
University of Hong Kong, Hong Kong,
People’s Republic of China)
[7] Evaluation of Huffman and Arithmetic
Algorithms for Multimedia Compression
Standards
Asadollah Shahbahrami,Ramin Bahrampour
,Mobin Sabbaghi Rostami,
Mostafa Ayoubi Mobarhan,
Department of Computer Engineering,
Faculty of Engineering, University of
Guilan, Rasht, Iran
[8] Data Compression Scheme of
Dynamic Huffman Code for Different
Languages
1Shivani Pathak, 2Shradha Singh
,3 Smita Singh,4Mamta Jain
, 5Anand Sharma
[9] Huffman Data Compression
Joseph Lee
©May 23, 2007